A lot of people have very good experience with analog in/out minidsp.
The reason is that ADC,DSP and DAC uses the same clock. So clock quality can be much lower. In/out thd + noise is around 90 dB. Good enough for all 92 or less sensituvity speaker elements.
The hiss sounded too pleasant to my ears, and it is silent while booting up (2x4 HD). It is more likely dither introduced for the Delta-Sigma converters to keep them happy while they have nothing to do. When there is sufficient signal level the dither can be dialed down, so the hiss is noticeable but not representative of the real performance.
It cannot be called audiophile grade equipment for this reason, but that was probably never the intention - the unit was designed to be useful, not designed for the audiophile.
I do wish the HD version comes available with balanced output. Unfortunately not.
It cannot be called audiophile grade equipment for this reason, but that was probably never the intention - the unit was designed to be useful, not designed for the audiophile.
I do wish the HD version comes available with balanced output. Unfortunately not.
So if there is a signal/music is playing then the hiss/dithering noise is not present to impact very soft passages of music ? Would the hiss/dithering noise come back if there a long enough period of "silence" ?
What you said is for a PCM DAC, which guarantees it is inaudible.
I am no expert and I may have abused the term "dither" but Delta-Sigma works differently from PCM. Intuitively Delta-Sigma is making up for the lack of bit resolution with a higher clock rate. So it follows that the noise floor (dither) will have to be higher when the converter has nothing to do.
The hiss noise I hear from miniDSP is certainly higher than what I hear from a vinyl record, and almost reaches the level of a metal cassette tape with Dolby C noise reduction.
I think every signal above approx -60dBFS will be able to completely bury the hiss. But -60dB is in itself clearly audible if you pay attention to it (such as in your period of silence example).
This has NEVER detracted from my own enjoyment in practice, but I can see where this is heading - a Bruckner symphony played in a Gothic cathedral playing a fermata. No musician is playing but you are listening to the tiny little reverberations dying down, and - here comes the miniDSP hiss...
Hello!
I've come to this thread because I'm undecided if committing to the minisharc dsp.
Pros are the I2S inputs and outputs, number of FIR taps, salve mode.
The only con is that my wav files are sampled at 44.1Khz so the minisharc will resample them to 48Khz before working with them.
In my opinion ASCR (asynchronous resampling) is very bad for sound quality because data samples are created out of thin air. So the question is: what is the alternative? What existing product or what combination of software and processor will let me perform convolution on my original, bit perfect data samples ?
Thanks
I've come to this thread because I'm undecided if committing to the minisharc dsp.
Pros are the I2S inputs and outputs, number of FIR taps, salve mode.
The only con is that my wav files are sampled at 44.1Khz so the minisharc will resample them to 48Khz before working with them.
In my opinion ASCR (asynchronous resampling) is very bad for sound quality because data samples are created out of thin air. So the question is: what is the alternative? What existing product or what combination of software and processor will let me perform convolution on my original, bit perfect data samples ?
ThanksHello!
I've come to this thread because I'm undecided if committing to the minisharc dsp.
Pros are the I2S inputs and outputs, number of FIR taps, salve mode.
The only con is that my wav files are sampled at 44.1Khz so the minisharc will resample them to 48Khz before working with them.
In my opinion ASCR (asynchronous resampling) is very bad for sound quality because data samples are created out of thin air. So the question is: what is the alternative? What existing product or what combination of software and processor will let me perform convolution on my original, bit perfect data samples ?
I also try to avoid unnecessary sample rate conversion, although if I'm going to then throw away all concept of 'bit perfect' with convolution I then become pretty relaxed about that rate conversion.
I'm not aware of any hardware based DSP that uses 44.1khz internally, maybe you might find a software based solution more easily if avoiding that rate conversion is the more important factor for you.
Hello thanks for your reply.
Well yes, once you perform convolution bit perfect flies away. Still I feel better working (perform convolution) with the original data set than a fabricated one.
I'm probably going to get a minidsp anyway and see how it performs. Then maybe replace the clock with a 44.1khz multiple...
Well yes, once you perform convolution bit perfect flies away. Still I feel better working (perform convolution) with the original data set than a fabricated one.
I'm probably going to get a minidsp anyway and see how it performs. Then maybe replace the clock with a 44.1khz multiple...
Well - I sort of hate to tell you, but once the dsp start doing calculations, your signal will be changed anyways...
Johan-Kr
That is understood.
The question is how the signal changes.
If you perform convolution (ie digital signal processing) on the original 16 bit data stream, 32bit floating point calculations will deal nicely with rounding errors. In the end you will have to round the numbers (again) to convert back to 16 bits, but you will get "one degree of error".
Going from 44.1Khz to 48Khz means guessing the second sample which is at a 0.91875 distance between the first two (44.1Khz) samples. (Good luck with that being correctly done). So in the end you will have "two degrees of error".
We could argue that conversion algorithms are very good at it but my idea is that it is best to remove what is not necessary.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.